基于H-K算法的MIDI文件主旋律音轨提取探讨  

Exploration of MIDI Files Based on H-K Algorithm Extract Melody Tracks

作　　者：马青阁 曹西征[1] 赵宛 汪旭彬 关金晨 

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《无线互联科技》2016年第6期65-66,共2页Wireless Internet Technology

基　　金：河南师范大学大学生创新创业训练计划资助项目;项目编号:20140005

摘　　要：主旋律是音乐旋律信息的重要组成部分,文章通过对表征音乐旋律特征向量的提取,采用H-K分类算法构建音轨分类器模型,对MIDI片旋律音轨和伴奏旋律音轨进行分类。最后通过候选音轨提取出主旋律音轨,研究表明,这种方法对10个音轨内的主旋律提取准确率由75%提升到90%,对于2个音轨以内的主旋律提取可以达到95%的准确率。Theme is an important part of music melody information, based on the characterization of music melody feature vector extraction, the H- K audio classification algorithm to construct classifier model, MIDI audio piece of melody and accompaniment melody track for classification. Finally through the candidate track extract melody track, research has shown that this approach within the 10 tracks on the main melody extraction accuracy increase from 75% to 75%, the two tracks of less than the main melody of extraction can reach 95% accuracy.

关 键 词：MIDI 主旋律 音轨熵 特征布尔数组 特征向量 

分 类 号：TN912.3[电子电信—通信与信息系统]